This change ensures that the native activity releases the KeyEvent
objects that it allocates since they will not be released
automatically and eventually the JNI slot table will overflow
and crash the process.
Bug: 3170494
Change-Id: I2ea527f056096fae9bc9e7685dccb56ec7a0e8b4
extract mode IME window
Make the SelectionModifierCursorController track its show/hide state
more accurately in the presence of an extract mode IME.
Lazily initialize CursorControllers so that we don't end up creating
them for various TextViews where they are not needed.
Change-Id: Ib76a95d9e7669016cf11b8b0add8843b438068ea
request a layout when asking the transparent region to be recomputed, without this
it may never happen, since the transparent region is only calculated during layout.
this also fixes an issue where some fullscreen apps will run slower than expected
because their transparent region is never processed.
Change-Id: Iba187ef7b23f08ad3454aaa6707a3d262551e30c
Add a method to the test utils which waits to accept a bonding request from
another device. This is identical to the pairing method except that it
waits for a remote pair request instead of initiating the pairing
request itself. Also, add a test method to the test utils.
Change-Id: I6391d03e9f33c040afd64d73083acfe3184251eb
The implementation was guarenteed to cause deadlock when a timeout was set.
Change-Id: I59444ea784eb9057c6c4c9e9123f558b3ef5eef6
Signed-off-by: Nick Pelly <npelly@google.com>
Adding more detail to the output logging to help track down issues, fixing some
download completed notification counter flakiness and making reboot test more
robust, and reducing the number of concurrent downloads in testMultipleDownloads()
to 10. After talking with Vasu, this is probably more appropriate as it is
closer to typical usage scenarios, and as a side effect should trim some
time from the test runs. Large numbers of downloads will be left for stress
testing.
Change-Id: Ie337cfe9b8d27299d70553e39c60e241ff3afe66
We were telling the app, but we weren't saving the data for ourselves.
This also tells them to re-scroll when there's a new size.
Bug: 3144373
Change-Id: I9d12b714119ff02dd7d7f1cfa997d8a44475b9e9
We are leaving enough API so that you can see when any Tag is discovered,
get its ID, and get its NDEF messages.
But for advanced use - creating tag connections and writing messages - we have
2 problems. Firstly a lot of the code is untested
(RawTagConnection.transceive()), or in some cases known not to work
(NdefTagConnection.write()). Secondly, there is still debate about how to
best expose information about Tags.
The set of data/methods exposed for a Tag changes completely depending on the
tag technology. There may be multiple sets of technology implemented in a
single tag. Tag A may have technology X and Y, Tag B may have technology Y
and Z. Furthermore, some NFC controllers will be not be able to use all
technologies, and so Android Device 1 may see technology X and Y on Tag A but
Android device 2 may only see technology X. So we have a pretty challenging
set of constraints to work under, and we are not convinced the current Tag and
NdefTag class is the best approach going forwards.
The Tag application should be able to remain unbundled, since it just needs to
get incoming NDEF Messages.
Change-Id: Ic09f094f33794e10f8d730fffe011c9a5957e0ac
Signed-off-by: Nick Pelly <npelly@google.com>
Retrieve the service again from ServiceManager on RemoteException.
Change-Id: Ie227b52019e7deafeab712af1addd6d957f7a8ee
Signed-off-by: Nick Pelly <npelly@google.com>
